The influence of substitution decisions made by national team coaches on final match outcomes at UEFA EURO 2024

Introduction: Coaches leading national football teams during championship tournaments make decisions about tactical substitutions of players in critical match phases. This may be an attempt to change or defend a favorable score. This study focused on the time of decision-making of forced and planned substitutions, considering its characteristics: neutral, offensive, and defensive. The point of analysis of the substitutions was the match outcomes at the time of the substitutions and the final result and impact of the substitution concerning the result.

Methods: A total of 51 matches played during the UEFA EURO 2024 football tournament were analyzed, during which 466 player substitutions were made. For the statistical analysis of the degree and strength of the relationship between the variables, the chi-square test, Cramer's V coefficient, and machine learning were used accordingly.

Results: 72% of coaches’ decisions to player substitutions resulted from the decision to change the team's tactics by changing the team's setup or the players’ positions. The most common negative (69%) or positive (61%) impact occurred from the substitution of a player after the 20th minute.

Discussion: The decision trees used in the analysis determined the most advantageous time periods for coaches to make decisions about substitutions. The highest substitution effectiveness rate is obtained when the substitution is made between 60 and 85 min, and the lowest is made between 45 and 60 min.

1 Introduction

The basic skill of a football coach is the ability to manage a team in response to changing situations on the pitch (1). The goal is known; it is to win the match or, in some tournament situations, to secure a draw. Additional objectives that may result from the tournament table may be to score a certain number of goals or lose by a certain goal difference. Such a complex arrangement of the main goal and intermediate goals places very specific demands on the coach in terms of managing changing situations (2). The accuracy of the player substitutions determines the changes in the team's performance (3–5). An ineffective substitution is a mistake that is practically impossible to fix. Any substitution of a substituted player reduces the number of possible tactical options in the final part of the match (6, 7). Substitutions are primarily about rebuilding the team's physical potential (8). Decisions to make substitutions have a different nature. One of them is the changed efficiency of running performance. There are cases in which coaches consciously place players in the team who are not prepared to play the entire match. This situation applies to all positions except the goalkeeper, although more often to forwards and defenders. No studies have been found that take into account the player's lack of preparation to participate in the entire match and the resulting time of his substitution. The running performance of players is an indicator that influences their perception by the coach in the context of substitution. Some positions are more physically demanding, which forces more frequent use of substitutions (9, 10). Trewin (11) identifies the differences in the running strategy of players participating in the entire match and substitutes, showing that substitutes from the last 20 min of the match are characterized by a significantly higher work intensity (distance covered per minute of the match, distance with high intensity, accelerations). Substitutes may use all-out tactics more often, knowing that they will only play for a relatively short time in the match. However, Trewin's (11) research has limitations resulting from the sample size and requires verification on a wider group. During championship matches, especially tournaments, mitigating the effects of fatigue is the primary reason for making substitutions, in addition to changing the tactical solution (4, 5, 12). Tactical changes usually occur after the first 45 min of the match. Decreases in players' performance are visible throughout the match, but the most common breakdown occurs after approximately 60 min of the match (13–15). This is when the second peak in the number of substitutions (apart from the start of the second half) is visible (16). The third peak occurs around the 80th minute of the match when introducing a new player, which is usually aimed at increasing the intensity of the game and surprising the opponent with a change in the team's tactics (17–19). One of the criteria for an effective substitution is to achieve or exceed the intensity of the substituted players (13, 20), indicating that players introduced in the second half of the match performed 25%–63% more intensive running and sprinting in the last 15 min of the match compared with players who were on the pitch from the beginning of the match. In addition, Carling et al. (3) indicate that substitutes are characterized not only by relatively higher total distance and maximal intensity distance but also by shorter recovery time between repetitions. Bradley et al. (5) indicate that substitutes cover a greater relative distance in high-intensity running (19.8–25.1 km/h) than that in players who played the full match or were replaced. Strategic changes during the match's second half led to a reduced fatigue effect on the entire team. This observation was confirmed by the research of Hills et al. (12). Myers (16) formulates a decision rule according to which, if the team is losing, the change should be made before the 58th minute, the second before the 73rd minute, and the third before the 79th minute. The effectiveness of this rule is at the level of 42.27%, and the lack of its application in only 20.52% of cases allowed for a favorable change in the result. Considering many factors accompanying the course of the match accompanies the coach's decision-making process when deciding to make a change. The substitution has an impact on the team's tactical formation (17). An offensive substitution increases the probability of not only scoring a goal but also conceding one. The observation by Schneemann and Deutscher (21) indicates that when the team is losing, the intensity of the substitutes' effort is lower than expected. Therefore, the effectiveness of the substitution from the point of view of the result at which it occurs may be different, and an unfavorable result reduces the effectiveness of the substitution. In addition to the two basic tasks accompanying the change, i.e., maintaining or increasing the intensity of the team's play and correcting tactics, numerous studies indicate several other factors. Raab et al. (22) draw attention to the psychological aspect. The coach's decisions to make a change are also influenced by the experience of the player being changed (23, 24), the position of the host of the match (25–27), and the way of refereeing (28, 29). The tactical effect of changes may also have another dimension, not only strictly football-related, realized in the game. The influence of contextual factors on the coach's decisions cannot be ignored (18). These include the match's location, which increases the probability of modifying the initial formation and increased pressure from the fans (30), the current result of the match, and the losing team deciding to change earlier than the winning team (18). The quality of the opponent's game (31, 32) or the importance of the competition (33) also influences decisions about changes in the team composition.

Previous studies indicate many variables influencing coaches' decisions regarding substitutions. The UEFA European Championship 2024 is a specific high-profile sporting event, in which the best teams in Europe participate. The role of the coach of these teams is also specific during the entire preparations and sports competitions. Here, as in the case of club teams, we do not have a coach working with players for many months, systematically building the team during daily training. In the case of the national team, this work is based primarily on match observations and team composition for a selected tactical construct. This is an unusual situation for a coach working with a team. Therefore, analyzing the impact of situational variables on the time and tactics of substitutions performed by national team coaches is a further expansion of knowledge on this subject.

The analysis of the time structure of substitutions and their effectiveness based on a literature review allows us to hypothesize that substitutions made by coaches between the 60th and 85th minute of the match during UEFA EURO 2024 are significantly more probability to have a positive impact on the match outcomes compared with substitutions made before the 60th minute or after the 85th minute.

2 Materials and method

2.1 Study design

The data for the analysis were obtained from the official UEFA EURO 2024 match reports. The research aims to present the coaching decisions and their consequences based on statistical tests and a predictive machine learning (ML) component to identify relationships between the variables, time interval information, the impact of substitution, and the frequency of player substitutions in football. This study includes data analysis from individual phases, i.e., the entire tournament, the group stage, and the knockout stage.

2.2 Sample

All men's UEFA EURO 2024 tournament matches were analyzed (n = 51) (Figure 1).

Figure 1

Figure 1. Classification of the analyzed UEFA European Championship 2024 matches.

2.3 Variables and instrument

The variables used were based on the analytical approach from Iglesias et al. (34), as well as their categories as follows:

1. Competition stage (CS): (1) group stage and (2) knockout stage.

2. Match status (MS): result at the time the coaches made any player substitution: (1) tying, (2) winning, and (3) losing.

3. Player substitution time (PST): minute in which the player substitutions were made:

a. (0) no player substitution;

b. (1) minute 0:00 to end of the first half;

c. (11) minute 45:00 to 50:00;

d. (12) minute 50:01 to 55:00;

e. (13) minute 55:01 to 60:00;

f. (14) minute 60:01 to 65:00;

g. (15) minute 65:01 to 70:00;

h. (16) minute 70:01 to 75:00;

i. (17) minute 75:01 to 80:00;

j. (18) minute 80:01 to 85:00;

k. (19) minute 85:01 to 90:00;

l. (20) minute 90:01 to end of the match.

4. Goal time (GT): minute in which a goal was scored. It presents the same categories as the PST variable.

5. Repercussion of the player substitutions on goals (RPSG): A classification was established that considers the period of time that elapsed from the player substitution until a goal was scored. This classification uses 5 min intervals (Figure 2). This variable was calculated as:

$\mathrm{RPSG}=\mathrm{GT}\text{-}\mathrm{PST}$

6. Impact of RPSG (IRPSG): (1) positive impact (own team goal after a player substitution), (2) negative impact (goal by the opposing team after a player substitution), and (3) no impact.

7. Final result (FR) of the match: (1) tie, (2) win, and (3) loss.

Figure 2

Figure 2. Interpretation of the RPSG variable (34).

2.4 The chi-square automatic interaction detection method

The process starts with collecting the data and selecting the ML model that best suits the aim of the study. Algorithms learn from the dataset and are then used to predict results or assess the current situation.

The chi-square automatic interaction detection (CHAID) method was used due to its properties in the analysis of categorical variables and its ability to detect complex interactions between multiple predictors through a hierarchical structure in the context of team management and coaching decisions.

As mentioned by Milanović and Stamenković (35), the decision tree is one of the most frequently used methods for effectively segmenting data based on statistically significant relationships, which aligns with our aim to identify combinations of match status, substitution timing, and final result that are most impactful.

2.5 Statistical analysis

All the statistical analyses were performed with OriginPro 2024 (OriginLab Corporation, USA). Statistical significance was set at α ≤ 0.05. For the analysis of the degree of relationship between the variables, the chi-square test (χ²) was used, and Cramer's V coefficient (Vc) indicated the strength of association between the variables. The strength of association was assessed by Crewson's guidelines, obtained value: <0.100 (small), 0.100–0.299 (low), 0.300–0.499 (moderate), and >0.500 (high) (36). After that, the adjusted standardized residuals (ASR) from the contingency tables were examined to detect patterns of association (37). Finally, the machine learning (ML) feature was applied to create the decision trees that were used to predict and identify interactions. The CHAID method was used to achieve interaction between variables by the academic open-source program Jeffrey's Amazing Statistics Program (JASP) (38).

3 Results

A total of 466 substitutions in the UEFA EURO 2024 tournament were analyzed, of which 328 player substitutions occurred during the group stage and 138 substitutions during the knockout stage (Table 1). The most common substitution according to player position was midfielder to midfielder (n = 126) of the entire tournament and individual phases (n = 92 and 34, respectively). A comparison of substitutions according to player position reveals a significant offensive advantage over the defensive (74 vs. 30) of the entire tournament. Similarly, the result is divided into individual phases: the group stage (48 vs. 18) and the knockout stage (26 vs 12).

Table 1

Table 1. Player substitutions including the player's position in the UEFA EURO 2024 tournament.

There are significant associations between the RPSG, FR, and IRPSG variables of the entire tournament with a moderate and high strength of association (Table 2). In individual phases, we can also notice the same results of significant associations between variables, except for the non-significant association in the knockout stage in short-term relationships with small strength. The no-repercussion relationship is statistically insignificant and was rejected in further analysis.

Table 2

Table 2. Degree of relationship between the RPSG, FR, and IRPSG variables of the UEFA EURO 2024 tournament.

A significant result of ASRs was observed in all matches where the final result was a win or loss. Still, a non-insignificant result was observed in matches ending in a tie of the entire tournament section (Table 3). In the group stage, ASRs were significant in wins or losses, except for immediate RPSG of defeat matches. All matches ending in a tie had insignificant ASR results. A significant result of ASRs was observed in the knockout stage, in immediate, medium-term, medium-long-term, and long-term time intervals where the final result was winning and in immediate, medium-term, and long-term time intervals of lost matches. A non-insignificant result was observed in all matches where the final result was a tie.

Table 3

Table 3. ASRs of the crossover of the RPSG, FR, and IRPSG variables of the UEFA EURO 2024 tournament.

Regardless of the type of RPSG, the positive impact implies more cases in matches that end in a win (106 cases), while the negative impact implies more cases in matches that end in a loss (101 cases) of the entire tournament. Comparing matches that result in a tie, the positive impact (43 cases) outweighs the negative impact (30 cases). In a percentage comparison, the positive impact associated with winning the game was between 52% and 65%, while the negative impact related to losing the game was between 54% and 78%. We observe a more positive impact of players' substitutions in winning matches made in short-term or long-term time intervals. The more negative impact of player substitutions in losing matches was made in the medium-term or long-term time intervals.

In the group stage, the positive impact, regardless of the type of RPSG, implies more cases in matches that end in a win (61 cases), while the negative impact implies more cases in matches that end in a loss (57 cases). In the comparison of matches that result in ties, there is the advantage of positive impact (38 cases) than negative impact (28 cases). In a percentage comparison, the positive impact associated with winning the game was between 48% and 64%, while the negative impact related to losing the game was between 45% and 69%. We observe that the most positive impact of player substitutions in winning matches was made in the long-term time interval. The more negative impact of player substitutions in losing matches was made in the long-term time interval.

In the knockout stage, the positive impact, regardless of the type of RPSG, implies more cases in matches that end in a win (45 cases), while the negative impact implies more cases in matches that end in a loss (44 cases). In the comparison of matches that result in a tie, there is the advantage of positive impact (five cases) than negative impact (two cases). In a percentage comparison, the positive impact associated with winning the game was between 57% and 83%, while the negative impact related to losing the game was between 25% and 100%. We observe that the most positive impact of player substitutions in winning matches was made in the long-term time interval. The more negative impact of player substitutions in losing matches was made in the long-term time interval.

Figure 3 shows the decision tree designed with RPSG, FR as the independent variable, and IRPSG as the dependent variable of the entire tournament. We observe that 45.84% of the neutral impact is from no-repercussion player substitution. The positive impact is 20.89% of won matches, and the negative impact is 20.70% of lost matches. In the tie match status, the substitution of players made in the immediate, short-term, medium-term, and long-term time intervals had a 10.54% positive impact, and a 2.03% negative impact in the medium-term and long-term time intervals.

Figure 3

Figure 3. The decision tree designed with RPSG, FR as independent variable, and IRPSG as dependent variable of the entire tournament.

There are significant associations between all variables of the entire tournament with a moderate or high strength of association (Table 4). We can also observe significant associations between all variables in the comparison group to the knockout stage with a moderate or high strength of association.

Table 4

Table 4. Degree of relationship between the MS, FR, and IRPSG variables of the UEFA EURO 2024 tournament.

The highest number of player substitutions, with a significant result of ASRs of the entire tournament, was observed in all matches, with IRSPG having no impact where teams were tying but the final result was a tie, with IRSPG having a positive impact where teams were winning but a final result was win, and with IRSPG having a negative impact where teams are losing but final results were loss (Table 5). The highest number of player substitutions with a significant result of ASRs was observed, with IRSPG having no impact on all final results of matches in the group and knockout stages.

Table 5

Table 5. ASRs of the crossover of the MS, FR, and IRPSG variables of the UEFA EURO 2024 tournament.

Figure 4 shows the decision tree results designed with MS, FR as the independent variable, and IRPSG as the dependent variable of the entire tournament. We observe a 6.10% positive impact of player substitutions where the match status was losing, but the final result was a win. The neutral impact is 20.33% when the match status was tying and the final result was a tie and 24.58% when the match status was winning and the final result was a win. The positive impact of tying matches, which ended with a win, is 8.13%, and the negative impact of tying or winning matches, which ended with a loss, is 10.54%. There was a 4.62% negative impact of player substitutions on winning match status, but the final results were tied.

Figure 4

Figure 4. The decision tree designed with MS, FR as independent variable, and IRPSG as dependent variable of the entire tournament.

There are significant associations between variables of the entire tournament, with matches in which the match status was tied having a moderate strength of association (Table 6). The matches in which the match status was winning or losing are statistically insignificant, but with a low strength of association. Significant associations exist between variables in the group stage, with matches in which the match status was tying or winning with a moderate strength of association. The matches in which the match status was losing are statistically insignificant, but with a moderate strength of association. There are significant associations between variables in the knockout stage, with matches in which the match status was tying with a high strength of association. The matches in which the match status was winning or losing are statistically insignificant, but with a moderate strength of association.

Table 6

Table 6. Degree of relationship between the MS, FR, and PST variables of the UEFA EURO 2024 tournament.

Throughout the entire tournament, when a team is tying, if a player substitution was made in 60:01–65:00 min, we observe that 40% of the matches ended in a win (Table 7). If the substitution was made before 60 min, 33%–37% of the matches ended in a tie, and 33%–66% of the matches ended in a loss. When a team is winning, if a player substitution was made between 55 and 90 min, we observed that 67%–94% of the matches ended in a win. If the substitution was made before 55 min, we observed three cases of the matches ending in a tie, and the winning team had a 6%–15% chance of the player's substitution ending in a loss. When a team is losing, if a player substitution was made from 45:00 to 55:00 min, we observed that 5%–40% of the matches ended in a tie. If the substitution was made between 45 and 90 min, we observed that 66%–94% of the matches ended in a loss. There were a few cases of player substitution, ranging from 4% to 16%, when a team that was losing made a change to the final result to win the match.

Table 7

Table 7. ASRs of the crossover of the MS, FR, and PST variables of the entire tournament.

In the group stage, when a team is tying, if a player substitution was made in 60:01–65:00 min, we observed that 47% of the matches ended in a win (Table 8). If the substitution was made after 65 min, we observed that 64%–83% of the matches ended in a tie, and if the substitution was made before 65 min, we observed that 25%–62% of the matches ended in a loss. When a team is winning, if a player substitution was made between 60 and 90 min, we observed that 66%–93% of the matches ended in a win. If the substitution was made between 45:00 and 50:00 min, we observed three cases of the matches ending in a tie, and if the substitution was made before 60 min, we observed two cases of the matches ending in a loss. When a team is losing, if a player substitution is made after 55 min, we observed that 60%–93% of the matches ended in a loss. There was a player substitution when the losing team made a substitution and won the match.

Table 8

Table 8. ASRs of the crossover of the MS, FR, and PST variables in the group stage.

In the knockout stage, when a team is tying, if a player substitution was made after 70 min, we observed that 25%–77% of the matches ended in a tie (Table 9). If the substitution was made before 70 min, we observed that 22%–54% of the matches ended in a win. If the substitution was made after 80 min, we observed that 33% of the matches ended in a loss. There was a situation when the winning team made a player substitution, and the final result was a tie or loss. When a team is losing, if a player substitution was made in 60–65 min, there were cases where matches ended in a tie. If the substitution was made after 45 min, we observed that 75%–88% of the matches ended in a loss. There were a few cases in which a player substitution changed the final result to win the match.

Table 9

Table 9. ASRs of the crossover of the MS, FR, and PST variables in the knockout stage.

Figure 5 shows the decision tree results designed with MS, PST as the independent variable, and FR as the dependent variable. We observe that 31.34% of all player substitutions occurred in matches where the final result was a win and 32.53% in matches where the final result was a loss. Moreover, 14.79% of substitutions from 65 min to the end of the match resulted in a tie. In a match where status was tying, 5.18% of matches with player substitution time from 0 to 60 min ended in a loss, and 16.27% of matches with player substitution time from 45 to 80 min ended in a tie.

Figure 5

Figure 5. The decision tree designed with MS, PST as independent variable, and FR as dependent variable of the entire tournament.

Figure 6 shows the decision tree results designed with PST as the independent variable and FR as the dependent variable of the entire tournament. We can observe the player substitutions made from 0 to 60 min (12.94%) and from 70 to 75 min (12.01%) accounted for matches that ended in a loss and those ade from 45 to 50 min (10.91%) and from 60 to 90 min (56.56%) accounted for matches that ended in a win. In the final result of tie matches, the player substitution time from 90 min to the end of the match was at 7.58%.

Figure 6

Figure 6. The decision tree designed with PST as independent variable and FR as dependent variable of the entire tournament.

4 Discussion

The analysis of the decisions made by coaches of the national teams participating in the UEFA 2024 tournament aimed to identify the factors determining the substitutions made in the research, which is the subject of this work. The nature of the tournament and its sports regulations practically eliminated the possibility of a draw or a controlled defeat. From the second stage, each match had to be resolved, which increased the teams' determination to win the match in regular time. Therefore, it can be assumed that the basic tactical assumption in each match was a victory, and as time passed, the determination increased among both the players and the coach. The basic assumption of the analysis was the current match result, which was assumed to be the basic determinant of decisions about the selection and substitutions of players made by coaches of the best national teams in Europe. An important element that may differentiate coaches' decisions about making substitutions may be the stage of the competition. Previous research did not consider this factor in the analysis of replacement football players. In the case of the UEFA EURO 2024 tournament, the first stage, i.e., the group stage, gave the team a chance to advance to the next round despite defeat. The knockout stage did not provide this possibility. The “all or nothing” rule forces coaches to respond decisively to the result (20). The studies of Rey et al. (19), Myers (16), and Del Corral et al. (18) indicate that the status of the match and its current score are closely related to tactical changes and substitutions made. A match that can't lose in a lead situation forces the coach to decide to strengthen the defense. In an unfavorable result, decisions are directed at taking control of the center zone of the pitch and strengthening the attack formation. In the second case, decisions about the first substitution are made earlier. In the UEFA EURO 2024 tournament, decisions about defensive substitutions in the first and second stages of the tournament were made at a rate of 2.3 and 2.7 per match, respectively. Therefore, the coaches strengthened the defense much more in the knockout stage compared with the group stage. Decisions about offensive changes were also made more frequently in the knockout stage compared with the group stage, 4.1 and 4.2 per match, respectively. The decision to strengthen the attack led to a tactical solution most often used to change the result by deciding the match or leading to a draw. Tactical changes made during the match are neutral, defensive, and offensive (18, 19, 39). It is the current match score that influences the decisions made by the coach (1, 20). Defensive substitutions increase the quality of the team's defensive formation and condense the team's defensive space. Offensive substitutions increase control of the opponent's defensive space and reduce the distance between teams. In total, decisions to strengthen defense or attack formations by substituting a player, which affected the change in the team's formation, which is important for these formations, were made in 339 cases out of 466 changes made in the entire tournament. Therefore, 72.74% of coaches’ decisions about player substitutions resulted from the decision to change the team's tactics toward more offensive (214 changes) or defensive (125 changes). In the match time structure during the UEFA EURO 2024 tournament, the time intervals in which coaches have so far most often made decisions about substitutions are 61–65 min, 71–75 min, and 81–85 min. They correspond to the time intervals of the most common substitutions determined in the studies by Bradley et al., Myers, and Rey et al. (5, 16, 19). At the same time, it has been shown that the match status has a strong impact on the decision-making time. In the knockout stage, almost twice as many substitutions as in the three typical time periods mentioned were made after the 90th minute of the match, i.e., during added time or extra time. These are tactical substitutions aimed at quickly changing the result, which at that moment may be unfavorable for both teams (tie), strengthen the defense (winning), or strengthen the attack (losing). These substitutions made by coaches are based on the current result, in which the decision-making process was highlighted in the studies of Lago, Bradley and Noakes, Myers, and Rey et al. (1, 4, 16, 19). The effectiveness of decisions made about substitutions was assessed using the IRSPG coefficient, which expresses the impact of substitution on the match result in 5 min periods, counted from the time of the substitution. Decisions made by coaches in the context of the time that passes between their decision-making and the team's reaction to the change are statistically significant in games ending with victory or defeat. Therefore, statistical analysis did not distinguish a period of the match in which the coach's decision was not statistically significant. From the point of view of the accuracy of decisions made, it is important to determine the structure of successful changes that brought the expected effect and unsuccessful changes after which a goal was lost. The results of the analysis indicate a slightly larger number of decisions that had a positive impact (n = 106) on the match result than negative ones (n = 101). Substitutions with a positive impact were effective for all time intervals counted from the time of the player substitution. The most common negative (69%) or positive (61%) impact occurred from the substitution of a player after the 20th minute.

In winning matches, substitutions made in all time intervals positively affected 52%–65% (106 cases). The negative effect had a different strength depending on the time that had passed since the change was made. Each time, the effect was statistically significant, with variable strength in individual time intervals, characterized by a waveform. The negative effect of the substitution was not temporary. Between 54% and 78% (101 cases) of the number of substitutions had a negative effect on the team that lost the match, while for the winning teams, the negative effect concerned between 9% and 23% (19 cases) of substitutions. Such a high difference between the effect of the substitution made by the coach indicates that when teams of similar high sports level compete, the correct (29.21%) or incorrect decision to substitution has a decisive impact on the final result (25.04%). In game tactics, the coach's decisions are often conditioned by the decisions made by the opposing team (16, 19). Therefore, substitutions are not made according to the 1:1 rule, i.e., replacement of players nominally playing in the same position. In the UEFA EURO 2024 tournament, there were 235 such substitutions in the group stage, which constituted 71.64% of all substitutions, while in the knockout stage, there were 104, which constituted 75.36% of all substitutions in the individual phases of the tournament. During the UEFA EURO 2024 tournament, a very common case was that the team defended a favorable result. One of the decisions made by the coach is to replace a defender or defensive midfielder with an offensive player. The incoming striker (in position no. 10) aims to attack the opponent high, making it difficult for them to play from the gate. This increases the amount of football action being moved further away from the home gate. Another solution is to strengthen the defense by replacing a midfielder or striker with a defender when the score is favorable. A favorable result for the own team forces the opponent to increase offensive actions, which are often carried out by crossing from the side sectors or half-spaces. Tactical changes also include reducing the opponent's advantage by making a “densifying” change in a specific sector of the field. The substitutions resulting from the need to increase or maintain the intensity of the game were drawn by Carling et al., Bradley and Noakes, Bradley et al., Pan et al., Hirotsu and Wright, Padrón-Cabo et al., and Lorenzo-Martínez et al. (3–5, 8, 17, 40, 41).

In the UEFA EURO 2024 tournament, in the group stage, most player changes were neutral, corresponding to the individual final results: 74 changes (24%), matches that ended in a tie; 61 changes (18.9%), matches that ended in a win (defensive); and 64 changes (19.51%), matches that ended in a loss (offensive). In the knockout stage, the most common type of substitution is neutral in matches with 23 substitutions ending in a tie, 25 substitutions ending in a win, and 25 substitutions ending in a loss. The reasons may be various, and decisions about substitution are made before the start of the match. The course of the competition indicates the optimal time for the coach to make such a decision. The most common reason is a decrease in the game's intensity, and the substitution is to maintain or increase it (5). Player injuries require immediate decision-making about substitution (12). Coaching decisions can be prepared in advance when a player enters the game with an injury and spontaneously when the injury results from participation in the game. Planned substitutions may involve changing the game's tactics and moving players on the pitch. This is an element of game quality management. Replacing a player on the pitch who is fully capable with another player in the same position in practice can be difficult or impossible. This is easier in the national team, but much more difficult in the club team. Then, other positions are rotated to maintain the quality of the team's play. This may require an additional substitution. Such a change in tactics is usually accompanied by double or triple substitutions. As noted in the works of Almeida et al. and Smith et al. (33, 42), an important effect of the substitutions is the team's mental sphere. These types of decisions are usually made by the coach in the event of a prolonged unfavorable score or a change in the score at the end of the match. In the UEFA UERO 2024 matches, 14 substitutions were made at the end of the match (last 5 min) in the group stage, which constituted 4.26% of the total substitutions, and in the knockout stage, 29% and 21%, respectively. Of these, 30 (9.14%) substitutions led to a change in the result in the group stage, and 25 (18.11%) in the knockout stage. The coach can use five players for substitutions in three breaks during the game and one additional break, which is the break between the first and second half of the match. This regulation limitation on substitution work requires management of breaks, in which, to use all players, there must be at least one double substitution. Using a double substitution during the break between the first and second half of the match gives the coach freedom to manage substitutions throughout the second half of the match. During the UEFA EURO 2024 tournament, this strategy was used 23 times in the group stage and 11 times in the knockout stage, which represented 7% and 8% of all substitutions in matches, respectively.

To verify various predictive models of making player substitutions, an efficient tool called machine learning (ML) was used (34, 43, 44). ML, using data on the time of the substitution, the result at which the substitution took place, scoring or losing a goal after the substitution, and the final result of the match, learns the relationships between them and estimates the effect of the decision made by the coach (45, 46). The use of ML was used to create decision trees, which are used when, as in the UEFA EURO 2024 tournament, there are many variants of decisions that involve risk. So far, ML has been used many times in team games research—ice hockey (47, 48), handball (49, 50), basketball (51), American football (52), and baseball (53, 54). ML has also been used in the analysis of decisions made in football (16, 55–57).

To determine the relationships of the UEFA EURO 2024 tournament match results, predictive analytical models and decision trees were used. The final result (FR) was used as a predictor (independent variable). The target (dependent) variables created three groups: (1) RPSG and IRPSG, (2) IRSPG and MS, and (3) MS and PT. Analysis of the decision tree of Group 1 showed that 46% of substitutions made during UEFA EURO 2024 had no impact on the result, and 21% were positive and contributed to the win. In 20% of cases, the team lost a goal after the substitution. Group 2 of dependent variables indicates that in 25% of cases, the substitutions made by the losing team are neutral and lead to no change in the result. In a similar percentage of cases, a substitution made by the winning team leads to another goal. In 20% of cases, substitutions made during a tie do not result in a change in the score. In 6% of cases, scoring a goal does not result in a win, and in 4% of cases, losing a goal after a substitution does not result in a loss. Group 3 dependent variables indicate that 31% of substitutions made by the winning or losing team do not influence the change of the match winner. No effect of changing the final result in the event of a tie is observed in 14% of substitutions made after the 65th minute of play. Research has shown that in the UEFA EURO 2024 tournament, nearly half (46%) of coaches' decisions to substitute did not affect the final result. An important issue for the coach, due to the effect of making a decision to substitution, is the predictable time of the potential effect of the substitution (4, 17, 18). The player reaches full effectiveness of action after a certain period of time has passed since joining the game (3, 5). During the UEFA EURO 2024 tournament, the most common period of time determining a positive or negative final result was the period above 21 min. The second significant time period that influenced the change in the final result was a positive effect between the 6th and 10th minute and a negative effect between the 11th and 15th minute. This results from the adaptation process to the intensity of the player himself and the remaining players from adapting tactics to the new position on the pitch (19). In the UEFA EURO 2024 tournament, only 8% of substitutions were effective within the first 5 min. The highest efficiency at the level of 20% is defined as medium-term after the substitution. Both the time interval and the nature of player substitutions played a significant role in the teams' efforts to win at the level of decisions made by the coach.

4.1 Limitations and future research

All matches of the UEFA EURO 2024 tournament were analyzed, but the relatively small number of matches (n = 51) limits the generalizability of the findings to other international competitions. The analysis relies solely on quantitative data, without considering the contextual or qualitative aspects of decision-making, such as individual coaching strategies, player readiness, communication during substitutions, changes in team tactics, changes in player performance during the game, and team cooperation on the pitch. The use of decision trees, while helpful in predictive modeling, may oversimplify dynamic game scenarios and cannot fully account for the non-linear nature of match outcomes.

The CHAID algorithm, while interpretable and appropriate for categorical analysis, does not include data on team formation or the reasons for forced substitutions, which may provide greater predictive accuracy. These factors should be considered when interpreting the model's results and in designing datasets for future studies.

Future research should aim to expand the dataset by including other international tournaments, domestic leagues, and women's competitions to assess whether similar patterns hold. Furthermore, the inclusion of kinematic data (e.g., total distance, high-speed running distance, and number of accelerations and decelerations) that indicate individual loadings on running performance could improve the assessment of the timing and effectiveness of substitutions in a match. Analyzing individual substitutions in the context of forced substitutions influences team tactics.

4.2 Study implications

The results of the study offer important insights for coaches and match analysts. The results suggest that changes made between the 60th and 85th minute are most likely to have a positive impact on match outcomes. Coaches should consider not only the running performance of players or changes in tactics but also the optimal periods for these changes to maximize their impact on results. The use of predictive models such as decision trees can support real-time decision-making tools in match management.

5 Conclusion

The UEFA EURO 2024 tournament is a collection of matches played in two systems—group and knockout, by national teams selected through European qualifications. The effects of decisions made by coaches during this tournament were partly different from those made in club teams' matches, in which previous analyses of this area of coaching work during the match were based. Earlier decision-making about substitutions is facilitated by the significantly higher quality of reserve players in national teams compared with club teams. The effectiveness of players' substitution toward the desired effect indicates the necessity of the coach's earlier reaction to the current result. This requires coaches to be willing to make player changes at the earlier stages of matches and adjust the initially adopted game strategy. The timing and nature of decisions made to make changes in the UEFA EURO 2024 tournament indicate that teams, after taking the lead, often decide to change their tactics and defend the result. Changing the result of a match as the game time runs out is difficult. Almost half of the changes do not affect the result. The most effective changes are those made in the medium-term time period. The least effective are those made in the last minutes of the match. Throughout the EURO 2024 tournament, effective substitutions that changed the match result toward a tie or win 5 min or less before the end of regular playing time were 17 and 20, respectively, which constituted 2.83% and 3.33% of all substitutions. Throughout the entire UEFA EURO 2024 tournament, only a few substitutions changed the result of the match from losing to victory. In almost half of the cases, coaches' decisions to substitute players have no impact on changing the match’s final result. The low effectiveness of changes toward reversing the match result indicates a limitation in the coach's ability to use this tool in conducting sports fights.

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions

TG: Conceptualization, Formal analysis, Methodology, Writing – original draft, Writing – review & editing. RC: Data curation, Formal analysis, Methodology, Visualization, Writing – original draft, Writing – review & editing. MG: Conceptualization, Formal analysis, Writing – original draft. MS: Data curation, Writing – original draft. KS: Data curation, Methodology, Writing – review & editing.

Funding

The authors declare that no financial support was received for the research and/or publication of this article.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declare that no Generative AI was used in the creation of this manuscript.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

1. Lago C. The influence of match location, quality of opposition, and match status on possession strategies in professional association football. J Sports Sci. (2009) 27:1463–9. doi: 10.1080/02640410903131681

PubMed Abstract | Crossref Full Text | Google Scholar

2. Black GM, Gabbett TJ, Johnston RD, Naughton G, Cole MH, Dawson B. The influence of rotations on match running performance in female Australian football midfielders. Int J Sports Physiol Perform. (2018) 13:434–41. doi: 10.1123/ijspp.2017-0175

PubMed Abstract | Crossref Full Text | Google Scholar

3. Carling C, Espié V, Le Gall F, Bloomfield J, Jullien H. Work-rate of substitutes in elite soccer: a preliminary study. J Sci Med Sport. (2010) 13:253–5. doi: 10.1016/j.jsams.2009.02.012

PubMed Abstract | Crossref Full Text | Google Scholar

4. Bradley PS, Noakes TD. Match running performance fluctuations in elite soccer: indicative of fatigue, pacing or situational influences? J Sports Sci. (2013) 31:1627–38. doi: 10.1080/02640414.2013.796062

PubMed Abstract | Crossref Full Text | Google Scholar

5. Bradley PS, Lago-Peñas C, Rey E. Evaluation of the match performances of substitution players in elite soccer. Int J Sports Physiol Perform. (2014a) 9:415–24. doi: 10.1123/ijspp.2013-0304

PubMed Abstract | Crossref Full Text | Google Scholar

6. FIFA. 2018 FIFA World Cup Russia™ (n.d.). Available online at: https://www.fifa.com/worldcup/ (Accessed December 21, 2024).

Google Scholar

7. Chmura P, Konefał M, Andrzejewski M, Kosowski J, Rokita A, Chmura J. Physical activity profile of 2014 FIFA World Cup players, with regard to different ranges of air temperature and relative humidity. Int J Biometeorol. (2017) 61:677–84. doi: 10.1007/s00484-016-1245-5

PubMed Abstract | Crossref Full Text | Google Scholar

8. Pan P, Li F, Han B, Yuan B, Liu T. Exploring the impact of professional soccer substitute players on physical and technical performance. BMC Sports Sci Med Rehabil. (2023) 15:143. doi: 10.1186/s13102-023-00752-x

PubMed Abstract | Crossref Full Text | Google Scholar

9. Di Salvo V, Baron R, Tschan H, Calderon Montero F, Bachl N, Pigozzi F. Performance characteristics according to playing position in elite soccer. Int J Sports Med. (2007) 28:222–7. doi: 10.1055/s-2006-924294

PubMed Abstract | Crossref Full Text | Google Scholar

10. Lago-Peñas C, Rey E, Lago-Ballesteros J, Casais L, Domínguez E. Analysis of work-rate in soccer according to playing positions. Int J Perform Anal Sport. (2009) 9:218–27. doi: 10.1080/24748668.2009.11868478

Crossref Full Text | Google Scholar

11. Trewin J. Match-running performance of elite female soccer players: the factors affecting performance and training applications (2018). Available online at: https://www.researchgate.net/publication/325930381 (Accessed October 5, 2024).

Google Scholar

12. Hills SP, Barwood MJ, Radcliffe JN, Cooke CB, Kilduff LP, Cook CJ, et al. Profiling the responses of soccer substitutes: a review of current literature. Sports Med. (2018) 48:2255–69. doi: 10.1007/s40279-018-0962-9

PubMed Abstract | Crossref Full Text | Google Scholar

13. Mohr M, Krustrup P, Bangsbo J. Match performance of high-standard soccer players with special reference to development of fatigue. J Sports Sci. (2003) 21:519–28. doi: 10.1080/0264041031000071182

PubMed Abstract | Crossref Full Text | Google Scholar

14. Reilly T, Drust B, Clarke N. Muscle fatigue during football match-play. Sports Med. (2008) 38:357–67. doi: 10.2165/00007256-200838050-00001

PubMed Abstract | Crossref Full Text | Google Scholar

15. Bradley PS, Sheldon W, Wooster B, Olsen P, Boanas P, Krustrup P. High-intensity running in English FA Premier League soccer matches. J Sports Sci. (2009) 27:159–68. doi: 10.1080/02640410802512775

PubMed Abstract | Crossref Full Text | Google Scholar

16. Myers BR. A proposed decision rule for the timing of soccer substitutions. J Quant Anal Sports. (2012) 8(1). doi: 10.1515/1559-0410.1349

Crossref Full Text | Google Scholar

17. Hirotsu N, Wright M. Using a Markov process model of an association football match to determine the optimal timing of substitution and tactical decisions. J Operat Res Soc. (2002) 53:1174–1174. doi: 10.1057/palgrave.jors.2601414

Crossref Full Text | Google Scholar

18. Del Corral J, Barros CP, Prieto-Rodríguez J. The determinants of soccer player substitutions. J Sports Econom. (2008) 9:160–72. doi: 10.1177/1527002507308309

Crossref Full Text | Google Scholar

19. Rey E, Lago-Ballesteros J, Padrón-Cabo A. Timing and tactical analysis of player substitutions in the UEFA Champions League. Int J Perform Anal Sport. (2015) 15:840–50. doi: 10.1080/24748668.2015.11868835

Crossref Full Text | Google Scholar

20. Hills SP, Barrett S, Feltbower RG, Barwood MJ, Radcliffe JN, Cooke CB, et al. A match-day analysis of the movement profiles of substitutes from a professional soccer club before and after pitch-entry. PLoS One. (2019) 14:e0211563. doi: 10.1371/journal.pone.0211563

PubMed Abstract | Crossref Full Text | Google Scholar

21. Schneemann S, Deutscher C. Intermediate information, loss aversion, and effort: empirical evidence. Econ Inq. (2017) 55:1759–70. doi: 10.1111/ecin.12420

Crossref Full Text | Google Scholar

22. Raab M, Bar-Eli M, Plessner H, Araújo D. The past, present and future of research on judgment and decision making in sport. Psychol Sport Exerc. (2019) 42:25–32. doi: 10.1016/j.psychsport.2018.10.004

Crossref Full Text | Google Scholar

23. Bar-Eli M, Tenenbaum G, Geister S. Consequences of players’ dismissal in professional soccer: a crisis-related analysis of group-size effects. J Sports Sci. (2006) 24:1083–94. doi: 10.1080/02640410500432599

PubMed Abstract | Crossref Full Text | Google Scholar

24. Mechtel M, Bäker A, Brändle T, Vetter K. Red cards: not such bad news for penalized guest teams. J Sports Econom. (2011) 12:621–46. doi: 10.1177/1527002510388478

Crossref Full Text | Google Scholar

25. Courneya KS, Carron AV. The home advantage in sport competitions: a literature review. J Sport Exerc Psychol. (1992) 14:13–27. doi: 10.1123/jsep.14.1.13

Crossref Full Text | Google Scholar

26. Carron AV, Loughhead TM, Bray SR. The home advantage in sport competitions: Courneya and Carron’s (1992) conceptual framework a decade later. J Sports Sci. (2005) 23:395–407. doi: 10.1080/02640410400021542

PubMed Abstract | Crossref Full Text | Google Scholar

27. Van Damme N, Baert S. Home advantage in European international soccer: which dimension of distance matters? Economics. (2019) 13(1):20190050. doi: 10.5018/economics-ejournal.ja.2019-50

Crossref Full Text | Google Scholar

28. Nevill AM, Balmer NJ, Mark Williams A. The influence of crowd noise and experience upon refereeing decisions in football. Psychol Sport Exerc. (2002) 3:261–72. doi: 10.1016/S1469-0292(01)00033-4

Crossref Full Text | Google Scholar

29. Dohmen TJ. The influence of social forces: evidence from the behavior of football referees. Econ Inq. (2008) 46:411–24. doi: 10.1111/j.1465-7295.2007.00112.x

Crossref Full Text | Google Scholar

30. Garicano L, Palacios-Huerta I, Prendergast C. Favoritism under social pressure. SSRN Electron J. (2001) 87(2):1–34. doi: 10.2139/ssrn.274934

Crossref Full Text | Google Scholar

31. Lago C, Casais L, Dominguez E, Sampaio J. The effects of situational variables on distance covered at various speeds in elite soccer. Eur J Sport Sci. (2010) 10:103–9. doi: 10.1080/17461390903273994

Crossref Full Text | Google Scholar

32. Bradley PS, Lago-Peñas C, Rey E, Sampaio J. The influence of situational variables on ball possession in the English Premier League. J Sports Sci. (2014) 32:1867–73. doi: 10.1080/02640414.2014.887850

PubMed Abstract | Crossref Full Text | Google Scholar

33. Almeida CH, Ferreira AP, Volossovitch A. Effects of match location, match status and quality of opposition on regaining possession in UEFA Champions League. J Hum Kinet. (2014) 41:203–14. doi: 10.2478/hukin-2014-0048

PubMed Abstract | Crossref Full Text | Google Scholar

34. Iglesias B, García-Ceberino JM, García-Rubio J, Ibáñez SJ. How do player substitutions influence men’s UEFA Champions League soccer matches? Appl Sci. (2022) 12:11371. doi: 10.3390/app122211371

Crossref Full Text | Google Scholar

35. Milanović M, Stamenković M. CHAID Decision tree: methodological frame and application. Economic Themes. (2016) 54:563–86. doi: 10.1515/ethemes-2016-0029

Crossref Full Text | Google Scholar

36. Crewson P. Applied Statistics Handbook, Version 1.2. Leesburg, VA, USA: AcaStat Software (2006).

Google Scholar

37. Williams C, Wragg C. Data Analysis and Research for Sport and Exercise Science. New York, NY, USA: Routledge (2003).

Google Scholar

38. Love J, Selker R, Marsman M, Jamil T, Dropmann D, Verhagen J, et al. Graphical statistical software for common statistical designs. J Stat Softw. (2019) 88(2):1–17. doi: 10.18637/jss.v088.i02

Crossref Full Text | Google Scholar

39. Gomez M-A, Lago-Peñas C, Owen LA. The influence of substitutions on elite soccer teams’ performance. Int J Perform Anal Sport. (2016) 16:553–68. doi: 10.1080/24748668.2016.11868908

Crossref Full Text | Google Scholar

40. Padrón-Cabo A, Rey E, Vidal B, García-Nuñez J. Work-rate analysis of substitute players in professional soccer: analysis of seasonal variations. J Hum Kinet. (2018) 65:165–74. doi: 10.2478/hukin-2018-0025

PubMed Abstract | Crossref Full Text | Google Scholar

41. Lorenzo-Martínez M, Rey E, Padrón-Cabo A. The effect of age on between-match physical performance variability in professional soccer players. Res Sports Med. (2020) 28:351–9. doi: 10.1080/15438627.2019.1680985

PubMed Abstract | Crossref Full Text | Google Scholar

42. Smith MR, Thompson C, Marcora SM, Skorski S, Meyer T, Coutts AJ. Mental fatigue and soccer: current knowledge and future directions. Sports Med. (2018) 48:1525–32. doi: 10.1007/s40279-018-0908-2

PubMed Abstract | Crossref Full Text | Google Scholar

43. Maneiro R, Casal CA, Ardá A, Losada JL. Application of multivariant decision tree technique in high performance football: the female and male corner kick. PLoS One. (2019) 14:e0212549. doi: 10.1371/journal.pone.0212549

PubMed Abstract | Crossref Full Text | Google Scholar

44. Gifford M, Bayrak T. A predictive analytics model for forecasting outcomes in the National Football League games using decision tree and logistic regression. Decis Analyt J. (2023) 8:100296. doi: 10.1016/j.dajour.2023.100296

Crossref Full Text | Google Scholar

45. Kabakchieva D. Predicting student performance by using data mining methods for classification. Cybernet Inform Technol. (2013) 13:61–72. doi: 10.2478/cait-2013-0006

Crossref Full Text | Google Scholar

46. Pandey M, Kumar Sharma V. A decision tree algorithm pertaining to the student performance analysis and prediction. Int J Comput Appl. (2013) 61:1–5. doi: 10.5120/9985-4822

Crossref Full Text | Google Scholar

47. Gliznitsa M, Silkina N. Using decision trees to determine the important characteristics of ice hockey players. Lect Notes Electr Eng. (2022) 857:359–69. doi: 10.1007/978-3-030-94202-1_34

Crossref Full Text | Google Scholar

48. Kim M-C. A study on the winning and losing factors of para ice hockey using data mining-based decision tree analysis. Appl Sci. (2023) 13:1334. doi: 10.3390/app13031334

Crossref Full Text | Google Scholar

49. Cabrera Quercini I, González-Ramírez A, García Tormo JV, Martínez I. Performance indicator selection through decision trees in elite handball. Revista Internacional de Medicina y Ciencias de la Actividad Física y del Deporte. (2022) 22:753–64. doi: 10.15366/rimcafd2022.88.003

Crossref Full Text | Google Scholar

50. Oytun M, Yavuz HU, Sekeroglu B. The performance analysis of male handball players using tree-based machine learning models. Int J Hum Move Sports Sci. (2024) 12:571–9. doi: 10.13189/saj.2024.120313

Crossref Full Text | Google Scholar

51. Horvat T, Havaš L, Srpak D. The impact of selecting a validation method in machine learning on predicting basketball game outcomes. Symmetry (Basel). (2020) 12:431. doi: 10.3390/sym12030431

Crossref Full Text | Google Scholar

52. Young CM, Luo W, Gastin P, Tran J, Dwyer DB. The relationship between match performance indicators and outcome in Australian football. J Sci Med Sport. (2019) 22:467–71. doi: 10.1016/j.jsams.2018.09.235

PubMed Abstract | Crossref Full Text | Google Scholar

53. Oh Y, Kim H, Yun J, Lee J-S. Using data mining techniques to predict win-loss in Korean professional baseball games. J Korean Institute Industr Engin. (2014) 40:8–17. doi: 10.7232/JKIIE.2014.40.1.008

Crossref Full Text | Google Scholar

54. Koseler K, Stephan M. Machine learning applications in baseball: a systematic literature review. Appl Artif Intell. (2017) 31:745–63. doi: 10.1080/08839514.2018.1442991

Crossref Full Text | Google Scholar

55. Schauberger G, Groll A. Predicting matches in international football tournaments with random forests. Stat Modelling. (2018) 18:460–82. doi: 10.1177/1471082X18799934

Crossref Full Text | Google Scholar

56. Baboota R, Kaur H. Predictive analysis and modelling football results using machine learning approach for English Premier League. Int J Forecast. (2019) 35:741–55. doi: 10.1016/j.ijforecast.2018.01.003

Crossref Full Text | Google Scholar

57. Min DK. Contribution analysis of scoring in the soccer game: using decision tree. J Korean Data Inform Sci Soc. (2019) 30:1385–97. doi: 10.7465/jkdi.2019.30.6.1385

Crossref Full Text | Google Scholar